northern hemispheric cryosphere response to volcanic...

Northern Hemispheric cryosphere response to volcanic eruptionsin the Paleoclimate Modeling Intercomparison Project 3 lastmillennium simulations

Mira Berdahl1 and Alan Robock1

Received 27 March 2013; revised 28 October 2013; accepted 29 October 2013; published 20 November 2013.

[1] We analyzed last millennium simulations (circa 850–1850 Common Era) from thePaleoclimate Modeling Intercomparison Project 3 (PMIP3) project to determine whethercurrent state-of-the-art models produce sudden changes and persistence of cold conditions afterlarge volcanic eruptions as inferred from geological records and previous climate modeling.Snow cover over Baffin Island in the eastern Canadian Arctic shows large-scale expansion(as seen in proxy records) in two of the five models with snow cover information available,although it is not sustained beyond a decade. Sea ice expansion in the North Atlantic is seen insome PMIP3 models after large eruptions, although none of these models produce significantcentennial-scale effects. Warm Baffin Island summer climates stunt snow expansion in somemodels completely, and model topography tends to miss the critical plateau elevations thatcould sustain snow on the island. Northern Hemisphere sea ice extent is lower in six of theeight models than in reconstructions over the past millennium. Annual average NorthernHemisphere mean climates have a range of 3K across models, while Arctic summer land-onlyclimates span more than 6K. This has critical consequences on ice and snow formation andpersistence in regions such as the Arctic where temperatures are near the freezing point andsmall temperature changes affect ice and snow feedback that could induce further climatechanges. Thus, it is critical that models accurately represent absolute temperature.

Citation: Berdahl, M., and A. Robock (2013), Northern Hemispheric cryosphere response to volcanic eruptions in thePaleoclimate Modeling Intercomparison Project 3 last millennium simulations, J. Geophys. Res. Atmos., 118,12,359–12,370, doi:10.1002/2013JD019914.

1. Introduction

[2] Earth’s temperature history over the past millenniumprovides a critical framework for accurately generating andunderstanding projections of future change. Northern Hemisphere(NH) proxy reconstructions over this period [Mann et al., 2009,2008; Jansen et al., 2007] have large uncertainties and showvariable timing, amplitude, and spatial extent of multidecadalevents such as the Little Ice Age (LIA, circa 1300–1850Common Era (C.E.)) and the Medieval Climate Anomaly (circa900–1250C.E.) [Frank et al., 2010;Miller et al., 2010]. Climatemodels are one of the most valuable tools used in the study ofthe last millennium, since they can help unravel the fundamentalunderlying mechanisms controlling the climate system.[3] Volcanism is the most prominent natural forcing factor

for climate variability of the last 700 years [Hegerl et al.,2007]. It is well known that strong volcanic eruptions (SVEs)can have a major impact on the global climate through

radiation impacts on the atmosphere [Robock, 2000], whichin turn impact ocean dynamics [e.g., Mignot et al., 2011;Miller et al., 2012]. SVEs tend to cool the global climatefor several years and can also cause regional effects suchas monsoon disruptions and winter warming over the NHcontinents [Robock, 2000].[4] Miller et al. [2012] presented data from ice caps on the

plateaus of Baffin Island in the eastern Canadian Arctic, show-ing intermittent ice melt and growth between the late thirteenthcentury and the mid-fifteenth century, followed by continuousice cover from the mid-fifteenth century until roughly acentury ago. The sudden cooling in the mid-thirteenth centurycoincides with four roughly decadally paced eruptions, thefirst of which is the 1257 Samalas eruption [Lavigne et al.,2013], the largest eruption in the last 7000 years [Langwayet al., 1988]. The snow expansion affected sites in north-centralBaffin Island in the elevation range of 660–1000m [Milleret al., 2012], dropping into the shallowly undulating plateausof 400–700m [Andrews et al., 1972]. Lowering the snowlineinto this region could cause a drastic increase in snow cover-age. Berdahl and Robock [2013] found that in a 10 km resolu-tion regional climate model, a summer temperature decreaseof 3.9 ± 1.1K from the current climate was necessary to lowerthe snowline by comparable elevation changes seen during thedescent into the LIA on Baffin Island.[5] Proxy records and some modeling experiments [Briffa

et al., 1998; Zhong et al., 2010;Miller et al., 2012; Schleussner

1Department of Environmental Sciences, Rutgers University, NewBrunswick, New Jersey, USA.

Corresponding author: M. Berdahl, Department of Environmental Sciences,Rutgers University, 14 College Farm Rd., New Brunswick, NJ 08901, USA.([email protected])

©2013. American Geophysical Union. All Rights Reserved.2169-897X/13/10.1002/2013JD019914

12,359

JOURNAL OF GEOPHYSICAL RESEARCH: ATMOSPHERES, VOL. 118, 12,359–12,370, doi:10.1002/2013JD019914, 2013

and Feulner, 2013] have suggested that it is possible to inducelong-term cooling given closely spaced multiple volcaniceruptions. Zhong et al. [2010] used the National Center forAtmospheric Research Community Climate System Model(CCSM3 model) at T42 resolution to test whether or not along-term cooling could be induced in the thirteenth centurywith transient volcanic perturbations. Under particular stabil-ity scenarios of the upper North Atlantic (NA) Ocean, the sim-ulations were able to induce abrupt snowline depressions fromdecadally paced explosive volcanism and sustained expandedsnow and ice conditions with sea ice/ocean feedbacks. Zhonget al. [2010] found that after the 1257 Samalas eruption, seasurface temperatures (SSTs) in the NA never recovered totheir pre-eruption temperatures, as several more large andclosely spaced volcanoes erupted and caused cumulativecooling. Cooler NA surface water was then advected to theAtlantic sector of the Arctic Ocean, reducing the rate of basalsea ice melt and allowing sea ice to remain in an expandedstate for more than 100 years.Mignot et al. [2011] found con-sistent results for post-thirteenth century eruptions with theirmodel, where they found a weakened Atlantic MeridionalOverturning Circulation and a sustained cooling and sea iceexpansion in the North Atlantic region. Schleussner andFeulner [2013] suggest that an increase in Nordic sea iceextent on decadal timescales as a consequence of major erup-tions leads to a spin-up of the subpolar gyre and a weakenedAtlantic Meridional Overturning Circulation, eventually caus-ing a persistent basin-wide cooling.[6] Here, we are interested in how the models from the

Paleoclimate Modeling Intercomparison Project Phase 3(PMIP3) last millennium (LM) simulations compare topaleoclimate reconstructions in the Arctic and NH.We assesswhether any of the models produce the reconstructed and pre-viously modeled sustained centennial-scale cold anomaliesand expanded sea ice and snow cover in the North Atlanticand Baffin Island regions following multiple, successive largevolcanic eruptions.

2. Methods

[7] Seven modeling groups participated in the LM inter-comparison project (Table 1). Each group participating ran

their coupled atmosphere-ocean model from 850 to 1850 C.E.,except Flexible Global Ocean-Atmosphere-Land System(FGOALS), which ran from 1000 to 1999 C.E. The modelswere forced with volcanic aerosol reconstructions from eitherGao et al. [2008] or Crowley et al. [2008], as outlined inTable 1. Both reconstructions are based on ice core sulfaterecords from both polar regions and differ in the transfer func-tion from the ice core sulfate to aerosol optical depth (AOD)and in the filtering of globally important eruptions [Schmidtet al., 2011]. There are two Goddard Institute for SpaceStudies (GISS) simulations, which differ predominantly intheir choice of volcanic forcing data set, hereafter referred toas GISSG, which used the Gao et al. [2008] reconstructionand GISSC, which used the Crowley et al. [2008] reconstruc-tion. Table 1 also notes the solar forcing used in each simula-tion. Schmidt et al. [2011] described the full details of climateforcing reconstruction options for the PMIP3 LM simulations.[8] Some models did not have snow or ice concentrations

available for download. Therefore, we are able to analyze nineunique models for temperature, of which one was GISS whichhad been run with two different volcanic forcing reconstruc-tions (GISSC and GISSG), seven models for sea ice, and fivemodels for snow.[9] Due to computational constraints, some of the models

were not able to properly spin-up and reach quasi equilib-rium. The Model for Interdisciplinary Research on Climate(MIROC) and GISS models show a drift in climate, the for-mer over the entire millennium and the latter over the first500 years. We address this in the GISS models by removingthe linear-fitted trend of the first 500 years of the control runfrom the forced simulations (GISSG and GISSC) [Brohanet al., 2012] and then adding the climatology of the rest ofthe control run (the last 500 years) back so we have absolutetemperature instead of anomaly. However, for MIROC, sincethe entire millennial record is drifting, we cannot return toabsolute temperature since we do not know the true climate.Thus, we leave MIROC uncorrected. The sea ice and snow inthe GISS simulations are corrected in a similar fashion, as isseen, for example, in Figure 5.[10] Typically, PMIP3 climate model results are compared

to proxy records as temperature anomalies, as opposed toabsolute temperatures [Brohan et al., 2012; Landrum et al.,

Table 1. Last Millennium Simulations Modeling Groups and Model Acronyms and Volcanic and Solar Forcing, Adapted FromBrohan et al. [2012]

Acronym Modeling Group Volcanic/Solar Forcing

BCC Beijing Climate Center, China Meteorological Administration [Wu et al., 2013] Gao et al. [2008]/Vieira and Solanki [2010]GISSG NASA Goddard Institute for Space Studies [Schmidt et al., 2006] Gao et al. [2008]/Vieira and Solanki [2010]GISSC NASA Goddard Institute for Space Studies [Schmidt et al., 2006] Crowley et al. [2008]/Vieira and Solanki, 2010FGOALS LASG, Institute of Atmospheric Physics, Chinese Academy of Sciences

[Bao et al., 2013; W. Man and T. Zhou, Simulated variability during thepast millennium: Results from transient simulation with the FGOALS-s2climate system model, Journal of Geophysical Research: Atmospheres,

under revision, 2013]

Gao et al. [2008]/Vieira and Solanki [2010]

MIROC Japan Agency for Marine-Earth Science and Technology, Atmosphere and OceanResearch Institute (The University of Tokyo), and National Institute for

Environmental Studies [Watanabe et al., 2010]

Crowley et al. [2008] / Delaygue and Bard [2011]and Wang et al. [2005]

MPI Max Planck Institute for Meteorology [Raddatz et al., 2007; Marsland et al., 2003] Crowley et al. [2008]/Vieira and Solanki [2010 ]and Wang et al. [2005]

CCSM4 National Center for Atmospheric Research [Gent et al., 2011] Gao et al. [2008]/Vieira and Solanki [2010]IPSL Institut Pierre-Simon Laplace [Marti et al., 2010] Gao et al. [2008]/Vieira and Solanki [2010]

and Wang et al. [2005]CSIRO The Commonwealth Scientific and Industrial Research Organization [Phipps, 2010] Crowley et al. [2008]/Steinhilber et al., [2009]HadCM3 Hadley Center [Collins et al., 2001] Crowley et al. [2008]/Steinhilber et al. [2009]

BERDAHL AND ROBOCK: PMIP3 CRYOSPHERIC RESPONSE TO VOLCANOES

12,360

2013]. While this method may be appropriate to evaluatevariability as relative responses to external forcing, in regionssuch as the Arctic where temperatures are near the freezingpoint (and phase changes occur), it is important to reportactual temperatures to evaluate the fundamental processesand states of the climate change. Thus, when reportingtemperatures in this article, we focus on absolute tempera-tures intentionally.[11] We compute area-weighted average temperatures for the

Arctic region (north of 66°N), the NA region, and Baffin Islandregion, shown in Figure 1. Unless otherwise noted (e.g., usingland only in these regions or Arctic defined as north of 60°Ninstead of 66°N), these are the regions referred to in the rest ofthis article. In some cases we apply a third-order low-passButterworth filter as inOtterå et al. [2010] so that we retain onlymultidecadal and slower variability for comparison to proxyreconstructions. To make the model results comparable to seaice reconstructions, sea ice extent in Figures 4 and 5 is calcu-lated as the areal sum of cells with ice concentration greaterthan 15%. If the grid cell has a sea ice concentration ≥15%,we consider it to be fully ice covered; otherwise, it is consid-ered ice free, in line with the National Snow and Ice DataCenter (NSIDC) sea ice extent definition.[12] Finally, we perform a superposed epoch analysis

(SEA) for our analysis of response to volcanic eruptions.We convert the Gao et al. [2008] NH sulfate aerosol loadingto aerosol optical depth (AOD) with the conversion factorthey recommend of 1.5 × 1014 g. We take the linear averageof the Crowley et al. [2008] AOD data for the 0°N–30°Nand 30°N–90°N latitude bands to generate a NH data setfor comparison to Gao et al. [2009]. We then find the top10 volcanic eruptions based on AOD in the NH for both datasets (Table 2). We superpose each eruption event and take theaverage of their response in temperature, sea ice area, andsnow cover area to each major eruption for the 25 years fol-lowing each event. Each parameter in the SEA analysis isreported as an anomaly with respect to the 5 year average

prior to the eruption. Sea ice in the SEA analysis is computedas an area as opposed to extent as in the other calculations inthis article. That is, no threshold of 15% is set; instead, wecompute the areal sum of the product of grid cell area withice concentration.

3. Results

3.1. Hemispheric, Arctic, and Regional Temperatures

[13] Figure 2 shows the average June-July-August (JJA)temperature series for the available PMIP3 LM simulationsfor the Northern Hemisphere, Arctic (north of 66°N), andBaffin Island (land only) region (270°E–300°E, 60°N–75°N).Since we are interested in decadal-scale variability, we useda third-order low-pass Butterworth filter with a cutoff fre-quency of 15 years [Otterå et al., 2010] to filter the tempera-ture records. The lowest panel shows the Gao et al. [2008]NH stratospheric loading of sulfates from volcanic eruptionsas an indicator of timing and magnitude of eruptions duringthe period. A prominent feature of all three temperaturepanels is the clear difference in mean states for each model,ranging 1.3 K, 2.8K, and 5.3 K in the NH, Arctic, andBaffin Island, respectively. The order of warmest to coldestmodel differs depending on the region of interest. OverBaffin Island, the average summer temperatures all tend tobe above freezing and those that dip below freezing inposteruption years are CCSM4 and GISSG.[14] Next, we compare the PMIP3 model results to paleo-

climate records in Figure 3. Since the temperature recon-structions are calibrated to instrumental records, there is nolarge uncertainty in their absolute calibration. All modelsexcept MIROC underestimate Arctic-wide JJA near-surfaceland temperatures. There is better representation by the modelsof NH annual temperatures, except for the Institut Pierre-Simon Laplace (IPSL) model which underestimates it byseveral degrees. The GISS models are particularly strong atrepresenting themean NH climate. These features are also seenin the zonal mean temperatures (not shown) for the NH meanannual and mean JJA temperatures for the full millennium.[15] The reconstructed temperature response to the eruptions

is generally muted compared to almost all of the modeledresponses (Figure 3). Timmreck et al. [2009] show that thetemperature response of the 1257 eruption (and presumablyall eruptions) is sensitive to the aerosol particle sizes prescribedin the simulation. They estimate the range of maximumsummer NH cooling after the 1257 eruption to be from 0.6K

Figure 1. Map showing Arctic region (north of 66°N) ingrey, North Atlantic (NA) region (270°W– 360°W, 50°N–90°N) in blue, and the Baffin Island region (270°W–300°W,60°N–75°N) in red.

Table 2. Top 10NHEruptions for theGao et al. [2008] andCrowleyet al. [2008] Data Sets Based on Aerosol Optical Depth (AOD)

Gao et al. [2008] Crowley et al. [2008]

Intensity Rank Year AOD Year AOD

1 1258 0.97 1258 0.662 1783 0.62 1816 0.403 1227 0.39 1809 0.294 1815 0.39 1641 0.285 1600 0.31 1600 0.266 1176 0.31 1228 0.187 1452 0.30 1460 0.158 1641 0.23 1668 0.159 939 0.21 1585 0.1410 1719 0.21 1024 0.14


12,361

to 2K. They show that only aerosol particle sizes substantiallylarger than observed after Pinatubo yield temperatureresponses consistent with reconstructions. The Gao et al.[2008] reconstruction contains no information about particlesize distribution, and the Crowley et al. [2008] data set

includes uncertain estimates of particle size, so using thewrong size distribution may be the reason that the tempera-ture response in the models to the 1257 eruption, and others,is overestimated. The magnitude of response in the Arcticand the NH after the 1257 eruption (Figure 3) is comparable

800 1000 1200 1400 1600 1800 200018

20

22NH

BCCGISSGGISSCCCSM4MIROCMPIIPSLCSIROFGOALSHadCM3

800 1000 1200 1400 1600 1800 2000−2

0

2

4Arctic

Tem

pera

ture

(°C

)

800 1000 1200 1400 1600 1800 2000

0

5

10Baffin

800 1000 1200 1400 1600 1800 20000

50

100

150

Year

NH

sul

fate

ae

roso

l (T

g)

Gao et al. [2008]

Figure 2. JJA temperature time series for each last millennium model simulation for the NorthernHemisphere, Arctic (north of 66°N) land and ocean, and Baffin Island (land only). Data are filtered witha third-order low-pass Butterworth filter with a cutoff frequency of 15 years. The bottom panel shows thetotal Northern Hemisphere stratospheric sulfate aerosol ejection (Tg) [Gao et al., 2008]. GISS models havebeen corrected for drift in the first 500 years.

1000 1200 1400 1600 1800 20002

4

6

8

10

Tem

pera

ture

(°C

)

Land−only Arctic Summer

BCCGISSGGISSCCCSM4MIROCMPIIPSLCSIROFGOALSHadCM3Reconst.GISSTemps

1000 1200 1400 1600 1800 200011

12

13

14

15

16

Year

Northern Hemisphere Annual

Figure 3. Comparison of paleoclimate temperature reconstructions to PMIP3 for the (top) Arctic summer(JJA) land-only north of 60°N and (bottom) NH annual temperature reconstruction. GISS models havebeen corrected for drift in the first 500 years. Arctic reconstruction is from Kaufman et al. [2009] andcompared to PMIP3 decadal averages. NH reconstruction is from Mann et al. [2009] and compared to10 year low-pass Butterworth filtered PMIP3 data. GISS temperatures are derived from land and oceanstation temperature anomalies for the NH from http://data.giss.nasa.gov/gistemp/. They were thenconverted to absolute temperature with the Jones et al. [1999] 1961–1990 base period average (14.6°C).


12,362

http://data.giss.nasa.gov/gistemp/

between the GISSG than the GISSC simulations, although thisis not the case for other major eruptions throughout the sim-ulations. Further error sources, such as location of the volcanoand season of eruption, complicate matters even more[Anchukaitis et al., 2012; Toohey et al., 2011]. In their assess-ment of the CCSM4 LM simulation, Landrum et al. [2013]found the response of the model to large eruptions to be 2–3times larger than the NH anomalies estimated from tree rings.The twentieth century response to volcanic eruptions is alsonoted to be too strong [Meehl et al., 2012]. Indeed, errors inthe paleoclimate reconstructions are very possible as well.Mann et al. [2012] suggested that NH volcanic responseswere being underestimated by tree ring-based reconstruc-tions, although there has been opposition to this suggestion[Anchukaitis et al., 2012].[16] Lastly, we assess whether or not the models are able

to reproduce a sustained Arctic LIA cooling. Proxy recordssuggest that NH temperatures decreased by roughly 0.5 K[Mann et al., 2009] and that at least a regional summer tem-perature lowering of 2 or 3K occurred in the North Atlanticsector of the Arctic [Miller et al., 2012]. We test whether themodels produce an Arctic and Baffin Island cooling withLIA period definition of 1450–1850 and 1600–1850. Wecompare this to the periods 850–1450 and 850–1600,respectively. We find that all models (except MIROC withthe LIA defined as 1450–1850) produce statistically signif-icant colder summer Arctic average temperatures (based ona Student’s t test), no matter which period is defined as theLIA. The maximum cooling is 0.3 K in the CCSM4 model.All other models produce less than 0.2 K change. We do thesame analysis for Baffin Island summer land temperaturesand find that four models produce statistically significantcolder temperatures in the LIA than in the period before.The maximum cooling achieved is again by the CCSM4model, of 0.5 K. Thus, the LIA is generated in most models,but its magnitude is very weak compared to what proxyrecords suggest.

3.2. Sea Ice

[17] Despite the issues with model resolution and snowcover representation on Baffin Island, another measure ofmultidecadal model response to repeated volcanic forcing isin the sea ice. Miller et al. [2012] cite an ocean/sea ice feed-back mechanism by which a centennial-scale expansion ofsnow cover and sea ice is maintained. Here, we assesswhether any of the PMIP3 models show a sustained sea iceexpansion in the North Atlantic region.[18] First, we compare the PMIP3 NH sea ice extent to the

Kinnard et al. [2011] Arctic August sea ice extent reconstruc-tion, which is based on 69 proxies predominantly derived fromice cores, but also from tree rings, lake sediments, and historicalobservations of sea ice. The reconstruction was calibratedagainst an historical index of late-summer extents (at least15% concentration) from 1870–1995. All PMIP3model ice ex-tents are calculated with a 15% concentration threshold as well,consistent with the paleoclimate reconstruction and NSIDCmethodology, where any grid cell with ≥15% ice concentrationis considered fully covered; otherwise, it is considered ice free.Aside from CCSM4 and the Beijing Climate Center (BCC)model, which largely fall within the reconstruction uncertaintyrange, the models tend to underestimate the sea ice extent forthe duration of the last millennium (Figure 4). The correlationof the models with reconstruction is usually significant( p values of< 0.05 using Student’s t test), but in seven of theeight runs, it is negative. MIROC shows the least sea ice area,although it is the only model to positively correlate with the re-construction. We did not find any consistent lags between indi-vidual models or between models and reconstructions.[19] The CCSM4 simulation shows a period of generally

higher sea ice extent between about 1250 and 1500 C.E.,suggesting that this model has a multidecadal sea ice responseafter the 1257 eruption. Figure 5 shows the minimum annualsea ice extent in the NA region for each model. The 40 yearlow-pass filtered time series exceeds the standard deviationof the low volcanic activity reference period (850–1150 C.E.)only in the GISSG, GISSC, and CCSM4 models. Thesignificant (>1σ) expansion of sea ice in these models lastsup to a few decades in GISSG and GISSC, and up to 80 yearsin CCSM4, but never reaches a continuous centennial longexpansion. Analysis of SSTs (not shown) in a subset of theNorth Atlantic (50°W–20°W and 50°N–65°N, per Zhonget al. [2010]) shows that none of the models produce a cumu-lative cooling after the 1257 Samalas and the subsequentclosely spaced eruptions. This was a key step in the mecha-nism for sustaining sea ice in the volcanically perturbedthirteenth century [Miller et al., 2012; Zhong et al., 2010].As a result, we infer that the rate of basal sea ice melt in theArctic Ocean was not reduced, and thus, sea ice did notremain in an expanded state for more than 100 years.

3.3. Model Elevation

[20] We suspect that model resolution plays a factor in prop-erly representing the mean climate over Baffin Island, whosehighest peak in reality, Mount Odin, reaches over 2 km abovesea level. Figure 6a shows the model representation and actualBaffin Island topography, along with the number of gridcells (n) comprising the Baffin Island region. Actual elevationsare derived from the U.S. Geological Survey GTOPO30 prod-uct, available from http://earthexplorer.usgs.gov/. This digital

500 750 1000 1250 1500 1750 20005

6

7

8

9

10

11

12

Year

Sea

Ice

Are

a (m

illio

n km

2 )

Figure 4. NH late summer (August) sea ice extent(millions km2) for PMIP3 simulations filtered with a third-orderButterworth 40 year low-pass filter compared to Kinnard et al.[2011] 40year smoothed reconstruction of sea ice extent basedon 69 proxy records. GISS models are corrected for first500years of drift. PMIP3models calculated using 15% ice coverthreshold (any cell ≥15% ice cover is considered fully covered,whereas anything <15% cover is considered open ocean).


12,363

http://earthexplorer.usgs.gov/

0 5

0 5

0 5

0 5 0 5

0 5

0 5

0 5

0 5

x 1011

0

500

x 1011

x 1011

x 1011

0

500

Ele

vatio

n (m

)

x 1011

x 1011

x 1011 x 1011 x 1011 x 1011

0

500

0

500

0

500

0

500

0

500

0

500

0

500

0 50

500

1000

1500

2000

0 1 2 3 4 5 6 7 8 90

500

1000

BCCGISSG GISSC

CCSM4

MIROCMPI

IPSLCSIROFGOALS

HADCM3

Max

imum

Mod

el

Ele

vatio

n (m

)

Mean JJA temp (°C)

0 2 4

x 109

1000

1500

2000

IPSLn=72

(b)

(a)BCCn =55 Observations

n=6485401

CCSM4n=400

GISSn=96

MIROCn=55

CSIROn=30

FGOALSn=99

HadCM3n=63

MPIn=136

Area (m2)

Figure 6. (a) Histograms of Baffin Island grid cell elevation distribution in terms of area represented foreach model and the observed digital elevation retrievals from the GTOPO30 data set (http://earthexplorer.usgs.gov/). The GTOPO30 observations panel inset emphasizes the high elevations present in reality onBaffin Island. Number of grid cells in the Baffin Island region is denoted with n. Few models representthe Baffin Island plateau elevation range of 400–700m, critical to snow expansion and persistence.(b) Maximum model elevation as a function of mean model summer (JJA) 2m temperature over BaffinIsland land from 850–1850 C.E. The red curve shows the best fit linear regression. There is a relationshipbetween maximum summer temperatures and the model topography of the island. GISS models have beencorrected for drift in the first 500 years.

−0.5

0

0.5

1 BCC

NA

Sea

Ice

Ext

ent (

106

km2 )

1000

1200

1400

1600

1800

1000

1200

1400

1600

1800

1000

1200

1400

1600

1800

CCSM4

1000

1200

1400

1600

1800

1000

1200

1400

1600

1800

1000

1200

1400

1600

1800

1000

1200

1400

1600

1800

1000

1200

1400

1600

1800

MIROC

−0.5

0

0.5

1 MPI 1 GISSG GISSC

UncorrectedCorrected

−0.5

0

0.5

1

−0.5

0

0.5

1

−0.5

0

0.5

1

−0.5

0

0.5

1

−0.5

0

0.5

−0.5

0

0.5

1CSIRO HADCM3

Year

Figure 5. Anomaly of minimum annual North Atlantic sea ice extent (grey) overlain with 40 year low-passfiltered series (black) with respect to the reference period 850–1150 C.E. (a time with low volcanic activity).Dashed lines show standard deviation of sea ice extent of reference. Models do not show centennial-scaleexpanded sea ice in this region after multiple closely spaced eruptions in the late thirteenth century. Sea iceextent is calculated with a 15% grid cell coverage threshold for full ice cover versus open ocean, and thedrift-corrected and uncorrected GISS curves are shown.


12,364



elevation model has a resolution of 30 arcseconds, whichroughly equates to 1× 1 km grids at the equator that get smallertoward the poles. The representation of elevations above 400mis deficient in most models compared to the GTOPO30 data(Figure 6a). None of the models represents elevations above750m, while in reality, more than 500 km2 of the region residesat higher elevations. The lack of representation of these highelevations in the models would not only have impacts on snowand ice, but also on atmospheric circulation. We examine therelationship between maximum model elevation and meansummer (JJA) surface temperature in Figure 6b. Models withhigher peak elevations on Baffin Island tend to have coldermean summer climates. The two models with the coldestclimate in the Baffin Island region, CCSM4, and GISS, alsorepresent the high plateaus (400–700m) [Andrews et al.,1972], elevations critical in fostering and sustaining largesnow area change [Berdahl and Robock, 2013].

3.4. Model Response to Volcanic Eruptionsin the Last Millennium

[21] The results of our SEA analysis for temperature, seaice, and snow cover with the top 10 eruptions in the lastmillennium are described below.3.4.1. Temperature[22] We analyze the response of the mean NH JJA and

annual temperatures as well as those over Baffin Island landonly (Figure 7). In all cases, the GISSG model shows thestrongest average peak response exceeding 1K NH annualaverage cooling and becomes even greater in the summermonths reaching roughly 1.5K.Allmodels show a post-eruptioncooling, MIROC being the most muted in the NH.[23] The temperature response on Baffin Island is again

strongest in the GISSG simulation, in both annual and sum-mer averages. Summertime average responses mainly governthe minimum snow cover extent on the island. Baffin Islandsummer temperatures cool by over 2K in the GISSG model,followed byGISSC andCCSM4. TheMIROC andBCCmodelsshow mild and slightly delayed post-eruption temperature

decreases compared to that of the GISS and CCSM4 models.The GISSG model shows the slowest recovery in the first 5post-eruption years, although the Max Planck Institute (MPI)model shows similar cooling in the decades following theeruption even though it does not reach such extreme coolingin year 1.[24] The integrated temperature response per unit forcing

from year 0 to year 10 is shown in Table 3 for NH andBaffin Island annual and JJA average responses for the top10 eruptions. The models with the largest integrated NHannual and JJA temperature response are MPI, HadleyCoupled Model (HadCM3), and the GISS model. On BaffinIsland, the integrated annual temperature response is strongestin MPI and the GISS models, whereas in the summer, theGISS, IPSL, and CCSM4 show the strongest net cooling inthe 10 years following the eruptions. Thus, the model temper-ature responses do not rank consistently from region to regionor season to season, and their impact on the cryosphere is notonly dependent on the temperature anomalies but also on themean state of the background climate.3.4.2. Sea Ice[25] The SEA analysis for September sea ice area in the NH

and the NA regions (Figure 8) follows that of temperature. Inthe NH, GISSG shows a significantly larger response than anyof the other models, more than twice the area anomaly thanany other model. All models subside to pre-eruption condi-tions by 10 years after the eruption. In the NA, GISSG showsthe largest peak in sea ice area anomaly after the eruptions, al-though it is followed closely by CCSM4, GISSC, and BCC.Again, the anomalies subside within roughly a 10 year lag ofthe eruption.[26] Integrated sea ice area expansion per unit forcing for

10 years following the eruptions is shown for the NH andNA in Table 3. In the NH, sea ice expansion is largest per unitforcing in the GISS models and HadCM3. In the NorthAtlantic, the GISS models and CCSM4 produce the greatestintegrated sea ice expansion in the 10 years following theeruptions. Again, over Baffin Island, the coldest models are

−2

−1

0

Tem

p. A

nom

aly

(K)

NH Annual

BCCGISSGGISSCCCSM4MIROCMPIIPSLCSIROFGOALSHadCM3

NH JJA

−4 0 4 8 12 16 20 24

−4 0 4 8 12 16 20 24 −4 0 4 8 12 16 20 24

−4 0 4 8 12 16 20 24

−2

−1

0

−2

−1

0

−2

−1

0

Lag (years)

Baffin Land Annual

Lag (years)

Baffin Land JJA

Figure 7. Superposed epoch analysis of NH annual, summer (JJA), Baffin Island (land only) annual, andJJA average temperature response to the top 10 eruptions in the last millennium. Anomaly is relative to the5 year average before eruptions. GISS models have been corrected for drift in the first 500 years.


12,365

CCSM4 and GISS, so it follows that they would produce themost sea ice in this region.3.4.3. Snow[27] The SEA analysis for snow is shown for the Baffin

region and the NH, expressed as a percentage of land coveredin snow at the annual minimum of monthly mean extent(typically August) and as a snow area anomaly with respectto the mean of 5 years prior to the eruptions (Figure 9). Theminimum annual percent of snow cover on land in the NHvaries across the models with CCSM4 differing fromMIROC NH snow cover by 3% (more than 150% change).In other words, the mean minimum annual snow extent in theMIROCmodel covers about 3.0 × 106 km2 less in area than thatof the CCSM4 model. As discussed earlier, this is probably inpart a function of model resolution since higher-resolutionmodels can capture higher elevations that are colder andsustain snow cover. The post-eruption snow extent anomalyin the NH is strongest and very similar in the GISSG and the

CCSM4 models. This response tapers off within a decade inall models, earlier in most.[28] On Baffin Island, the story is similar. The MIROC

model shows minimal snow retention at the peak of summeron the island and no significant post-eruption anomaly evenafter the eruptions in this region. The BCC and MPI modelsproduce a similar result, although their minimum snowextent tends to be higher. The CCSM4, GISSG, and GISSCmodels show significant responses to the eruptions on BaffinIsland, both in terms of percentage and area anomaly. TheCCSM4 and GISSG models show the same maximum areaanomaly over Baffin Island, roughly twice the magnitude ofthat seen in GISSC. Similar to the NH, the post-eruption snowexpansion is short-lived, decaying back to pre-eruption snowextents within 10 years.[29] Table 3 shows the NH and Baffin Island snow area

anomalies following the top 10 eruptions integrated over lagyears 0 to 10. In both cases, the snow expansion is largest in

−4 −2 0 2 4 6 8 10 12 14 16 18 20 22 24−0.5

0

0.5

1

1.5

2

2.5

3

Lag (year)

−4−2 0 2 4 6 8 1012141618202224

Lag (year)

Sea

Ice

Are

a A

nom

aly

(106

km2 )

Northern HemisphereBCCGISSGGISSCCCSM4MIROCMPICSIROHadCM3

−0.2

−0.1

0

0.1

0.2

0.3

0.4

0.5

0.6North Atlantic

Figure 8. Superposed epoch analysis for top 10 eruptions in the last millennium (Table 2), showingNorthern Hemisphere and North Atlantic September sea ice area. Anomaly is with respect to the meanof 5 years prior to the eruptions. Note on the different scales on the y axis. GISS models are corrected forthe first 500 years of drift.

Table 3. Temperature, Sea Ice, and Snow Anomaly Responses Per Unit Forcing (UF) Integrated From Year 0 to Year 10 Lag Afterthe Eruption

Model

NH AnnualTemperature(K/UF)

NH JJATemperature(K/UF)

Baffin AnnualTemperature(K/UF)

Baffin JJATemperature(K/UF)

NH Sea Ice Area(×106 km2/UF)

NA Sea Ice Area(×106 km2/UF)

NH Snow Area(×106 km2/UF)

Baffin Snow Area(×106 km2/UF)

BCC �3.5 �2.6 �8.4 �12.4 16.0 3.4 2.0 0.1GISSG �10.4 �10.2 �25.1 �40.9 24.1 5.3 9.8 3.0GISSC �10.2 �9.6 �57.6 �48.4 20.8 4.8 5.5 1.9CCSM4 �8.4 �8.2 �35.1 �18.4 7.1 3.5 9.5 3.1MIROC �5.0 �4.1 �3.8 �7.6 13.5 1.8 3.3 0.0MPI �11.9 �10.9 �1.5 �49.9 14.8 1.8 0.2 0.0IPSL �6.4 �6.2 �51.1 �28.1CSIRO �7.2 �6.3 �9.6 �10.9 10.5 2.1FGOALS �6.4 �5.7 �15.5 �6.0HadCM3 �10.7 �10.2 �20.9 �20 16.8 2.6

aAverage forcing for the top 10 eruptions in terms of aerosol optical depth (Table 2) is used to compute response per UF. Anomalies are calculated from the5 year pre-eruption climatology. A 10 year pre-eruption climatology was tested as well but made minimal difference to the results.


12,366

the GISS and CCSM4 models. Over Baffin Island, the largestintegrated response is generated by the CCSM4 and theGISSG model. Again, these are the coldest models in thisparticular region, although these models rank only third and

fourth in their summer temperature response to theeruptions over Baffin Island. This emphasizes that both thebackground summer climate and the strength of theresponse to volcanic eruptions simulated by the models

−4−2 0 2 4 6 8 1012141618202224 −4−2 0 2 4 6 8 1012141618202224

−4−2 0 2 4 6 8 1012141618202224−4−2 0 2 4 6 8 1012141618202224

0

20

40

60

% L

and

Cov

ered

in S

now

Baffin Island

−1

0

1

2

3x 105

x 105

Baffin Island

Sno

w A

rea

Ano

mal

y (k

m2 )

1

2

3

4

5

6

Lag (year) Lag (year)

Northern Hemisphere

−5

0

5

10

15Northern

HemisphereBCCGISSGGISSCCCSM4MIROCMPI

Figure 9. Superposed epoch analysis for the top 10 eruptions in the last millennium (Table 2). Panelsshow snow area coverage in the Baffin Region (270°W–300°W, 60°N–75°N) and for the NorthernHemisphere expressed as the percent of the total land area covered in snow at the minimum annual extentin the left panels. The right panels show the snow area anomaly with respect to the mean of the 5 years priorto the eruptions for the same regions. Lag of 0 represents the year of the eruption. Note on the differentscales on the y axes. GISS models have been corrected for drift in the first 500 years.

−2.5 −2 −1.5 −1 −0.5 0 0.5−1

0

1

2

3x 105

Min

imum

Ann

ual

Sno

w A

nom

aly

(km

2 )

JJA Temperature Anomaly (°C)

0 1 2 3 4 5 6−15

−10

−5

0

5x 104

Mean JJA Climate (°C)

Δ M

in. S

now

Are

a /

Δ JJ

A T

emp

(km

2 /°C

)

BCCGISSGGISSCCCSM4MIROCMPI

(a)

(b)

Figure 10. SEAminimum annual Baffin Island snow cover anomaly as a function of SEA JJA temperatureanomaly for the 10 years following the top 10 eruptions (lag years 1–10 in Figures 9 and 7, respectively).Anomalies are with respect to the SEA values for the 5 years prior to the eruptions (lag years �5 to �1 inFigures 9 and 7). GISS models have been corrected for the first 500 years of drift. (a) Ordinate value iscalculated as the slope for each model. (b) Magnitude of snow cover expansion on Baffin Island duringsummer cooling after the top 10 eruptions, plotted against the mean JJA Baffin Island climate for each model.Error bars show 95% confidence level. GISS models previously corrected for the first 500 years of drift.


12,367

dictate the behavior of the cryosphere. This is in agreementwith the general conclusions of Zanchettin et al. [2013].3.4.4. Model Sensitivity[30] Next, we assess the SEA temperature and snow

responses together, by plotting the temperature and snowSEA time series against each other (Figure 10a). We truncatethe time series to include only lag year 1 to lag year 10 sincewe are interested in post-eruption response. From this we canfit least-squares regression fits to each model and find thesensitivity of the snow response to temperature change foreach model (units of km2/K). Plotting these sensitivitiesagainst the mean JJA climate in each model over BaffinIsland, we show that the models’ ability to produce a snowresponse is strongly a function of the mean climate(Figure 10b). There is a sudden transition between modelsexhibiting a widespread change in snow cover and those withvirtually no change in snow. Above about 2.5°C mean sum-mer climate, the models become incapable of reaching thefreezing point and forming snow, let alone sustaining extraamounts of snow on Baffin Island. Below roughly 2.5°C,the models show widespread snow expansion. The temp-erature response after the eruption is also a function of thevolcanic reconstruction used, which is evident in the GISSmodels, whose primary difference is the choice of volcanicforcing reconstruction. This result highlights the necessityfor models to not just correctly reproduce eruption responseas anomalies, but equally important, they must be able tocapture the absolute temperature. This has particularly criti-cal consequences around the freezing mark, as the presenceof snow and ice triggers the ice-albedo feedback, one of themost important climate feedbacks on the globe.

4. Discussion

[31] We can summarize our main findings as follows.Comparison of NH, Arctic, and Baffin Island temperaturein the PMIP3 models shows a large spread in mean modelbackground climate, the rank of which is not consistentacross regions. Annual average NH mean climates span3K, Arctic average summer temperatures span 3K, andArctic land-only summer climates range by more than 6K.Over Baffin Island, summer temperatures dip below 0°C aftermajor eruptions only in the GISSG, GISSC, and CCSM4 sim-ulations, allowing snow to expand on the island only in thesemodels despite other models having a greater integrated tem-perature response per unit forcing. Mean summer climateof Baffin Island is partly a function of the model resolutionand grid cell elevations. Most models produce a Little IceAge; however, the magnitude of associated cooling is muchless than expected from reconstructions. The largest changenoted was about 0.5 K cooling on Baffin Island, comparedto the proxy data which suggests that at least several degreesof cooling occurred [Miller et al., 2012]. The temperature re-sponse to volcanic eruptions is generally stronger in the PMIP3models than in reconstructions, consistent with other studies[Landrum et al., 2013; Brohan et al., 2012; Meehl et al.,2012]. This could be a result of, for example, inadequate aero-sol representation [Timmreck et al., 2010], model difficulty incapturing dynamic responses in the stratosphere [Shindellet al., 2003], uncertainties in eruption location and time ofyear [Anchukaitis et al., 2012; Toohey et al., 2011], andpotential errors in ice core interpretation when generating

volcanic reconstructions [Schneider et al., 2009]. Problemswith the paleoclimate reconstructions are of course possibleas well. It has been suggested that there is an underestima-tion of volcanic cooling in tree ring-based reconstructionsof Northern Hemisphere temperature [Mann et al., 2012],although this has been vehemently rebutted by others[Anchukaitis et al., 2012].[32] Sea ice extent tends to be underestimated in most of the

models compared to the Kinnard et al. [2011] reconstruction.CCSM4 andBCC dowell in their sea ice extent representation,falling within the range of uncertainty of the reconstruction.Compared to the period 850–1150 C.E. which is characterizedby low volcanic activity, 40 year low-pass filtered sea icearea in the NA shows decadal-scale expansion occasionallyafter the mid-thirteenth century in the GISSG, GISSC, andCCSM4 models, but does not show centennial-scale expan-sion in the LM simulations. The key event in the Milleret al. [2012] modeling was volcanic activity in the late thir-teenth century. At that time, eruptions were closely spacedenough that the North Atlantic surface waters never recoveredto their pre-eruption temperature between eruptions, so thatthere was a cumulative temperature lowering larger than forany single volcanic event. Our analysis of SSTs in thePMIP3 models (not shown) in a subdomain of the NorthAtlantic (50°W–20°W and 50°N–65°N, per Zhong et al.[2010]) do not produce a sustained cooling beyond about a de-cade after the 1257 Samalas eruption. The relatively warmSST conditions in the PMIP3 simulations likely contribute tolimiting the sea ice expansion during the thirteenth century.[33] The SEA analysis reveals the temperature response to

SVEs in the NH is strongest in the GISSG model, with morethan 1K maximum annual average cooling and up to 1.5Kmaximum summer average cooling. MIROC and BCC showless than 0.5K maximum cooling for both annual and sum-mer cooling. Cooling is much stronger over Baffin Islandland, particularly in the summer, when GISSG exceeds 2Kcooling, and greater than 1K cooling is produced in all othermodels except BCC and MIROC. Minimum annual sea icearea expands most in GISSG, and this behavior is especiallyoutstanding from the other models in the NH compared to theNA. On average, CCSM4 produces the most expansive min-imum annual NH snow cover, followed by the BCC, GISS,MPI, and MIROC models. The only models that show a re-sponse in the SEA snow analysis in the NH and over BaffinIsland are GISSG, CCSM4, and GISSC models. OverBaffin Island, the largest integrated response in snow areafrom year lag 0 to 10 is in CCSM4 and GISSG. Again, theseare the coldest models in this region, although these modelsrank only third and fourth in their integrated summer temper-ature response to the eruptions over Baffin Island. Thus, acombination of appropriate background climate and responseto volcanic eruptions is necessary to generate a response inthe cryosphere, in agreement with the findings inZanchettin et al. [2013]. From the SEA snow and temp-erature analyses over Baffin Island, we find that the onlymodels that manifest a snow expansion in response to a tem-perature drop after SVEs are the models whose mean summerclimates are near enough to 0°C, such that a perturbationbrings temperatures below freezing. There is a sharp transi-tion between the degree of snow expansion in models witha mean climate below and above about 2.5°C. Zhong et al.[2010] and Miller et al. [2012] used CCSM3, CCSM4’s


12,368

predecessor, to produce a centennial-scale climate changefrom decadally paced SVEs. Here we show that the coldestmodel in the Baffin Island region is CCSM4, so it is likelythat CCSM3 also produces near-zero temperatures in this re-gion. It is possible, then, that they may not have found signif-icant and sustained sea ice and snow expansion if they hadused a climate model which happened to model a warmermean state in this region. Ultimately, we show that it is criti-cal to accurately model the absolute temperature in regionsnear the freezing point where snow and ice form and melt.Crossing this threshold could potentially induce feedback thatsupport further sea ice expansion, colder SSTs, and conse-quently colder and snowier conditions on land.

5. Conclusions

[34] We assessed the PMIP3 LM simulations in terms ofabsolute temperatures and the temperature, snow, and iceresponse to volcanic eruptions with a focus on the NorthAtlantic and eastern Canadian Arctic regions in order todetermine if current state-of-the-art models produce suddenand persistent cold conditions after SVEs. We have shownthat the PMIP3 models are generally colder than reconstruc-tions over Arctic land but at the same time have too little seaice in the Arctic compared to reconstructions. Most modelsproduce significantly cooler temperatures in the Arctic andin the North Atlantic region during the LIA; however, themagnitude of cooling is much less than expected from proxyrecords. Only two of the models produce decadal sea iceexpansion, but none produce centennial-scale expansion. Snowcover over Baffin Island shows large-scale expansion in onlytwo of the five available models, although it is not sustainedbeyond a decade. The PMIP3 models’ lack of sustainedcooling response could be due to their inability to capturechanges in ocean circulation, which other studies have showncan lead to centennial-scale cooling in the North Atlantic.[35] Spread in the models’ mean climate states in the NH,

Arctic, and over Baffin Island is evident. Model resolution,and consequently topography, plays a strong role in determin-ing mean climate over Baffin Island. Critical plateau elevationsof 400–700m, necessary for fostering and sustaining largesnow area change [Berdahl and Robock, 2013], are only repre-sented in the CCSM4 andGISSmodels. Inmore than half of themodels, warm summer climates over Baffin Island stunt snowexpansion after volcanic eruptions completely. Thus, it is cru-cial to properly represent absolute temperatures particularly inareas such as the Arctic where small temperature changes dic-tate phase changes of water. This is especially important, sincesnow and ice presence can further induce feedback, such as ice-albedo, which may influence global climate but cannot be trig-gered from volcanic forcing in models that are too warm.

[36] Acknowledgments. We thank Giff Miller, Davide Zanchettin,and an anonymous reviewer for their thoughtful comments on this paper,which very much improved this manuscript. We thank Philip Brohan andJohann Jungclaus for sharing valuable insight regarding the PMIP3 modelsand Reto Ruedy for help with GISS temperatures. We acknowledge all ofthe modeling groups in the PMIP project for producing and making availabletheir results through the Earth Systems Grid Federation archive. This workwas supported by NSF grants ARC-0908834 and AGS-1157525.

ReferencesAnchukaitis, K. J., et al. (2012), Tree rings and volcanic cooling,Nat. Geosci.,5(12), 836–837.

Andrews, J. T., R. G. Barry, R. S. Bradley, G. H. Miller, and L. D. Williams(1972), Past and present glaciological responses to climate in easternBaffin Island, Quat. Res., 2(3), 303–314.

Bao, Q., et al. (2013), The flexible global ocean-atmosphere-land systemmodel, spectral version 2: FGOALS-s2, Adv. Atmos. Sci., 30, 561–576.

Berdahl, M., and A. Robock (2013), Baffin Island snow extent sensitivity:Insights from a regional climate model, J. Geophys. Res. Atmos., 118,3506–3519, doi:10.1002/jgrd.50249.

Briffa, K. R., P. D. Jones, F. H. Schweingruber, and T. J. Osborn (1998),Influence of volcanic eruptions on Northern Hemisphere summer temper-ature over the past 600 years, Nature, 393, 450–455, doi:10.1038/30943.

Brohan, P., R. Allan, E. Freeman, D.Wheeler, C.Wilkinson, and F.Williamson(2012), Constraining the temperature history of the past millennium usingearly instrumental observations, Clim. Past, 8, 1551–1563, doi:10.5194/cp-8-1551-2012.

Collins, M., S. F. B. Tett, and C. Cooper (2001), The internal climate variabilityof HadCM3, a version of the Hadley Centre Coupled Model without fluxadjustments, Clim. Dyn., 17(1), 61–81.

Crowley, T. J., G. Zielinski, B. Vinther, R. Udisti, K. Kreutzs, J. Cole-Dai,and E. Castellano (2008), Volcanism and the Little Ice Age, PAGESNewslett., 16, 22–23.

Delaygue, G., and E. Bard (2011), An Antarctic view of Beryllium-10 andsolar activity for the past millennium, Clim. Dyn., 36, 2201–2218,doi:10.1007/s00382-010-0795-1.

Frank, D. C., J. Esper, C. C. Raible, U. Buentgen, V. Trouet, B. Stocker,and F. Joos (2010), Ensemble reconstruction constraints on the globalcarbon cycle sensitivity to climate, Nature, 463, 527–530, doi:10.1038/nature08769.

Gao, C., A. Robock, and C. Ammann (2008), Volcanic forcing of climateover the past 1500 years: An improved ice core-based index for climatemodels, J. Geophys. Res., 113, D23111, doi:10.1029/2008JD010239.

Gent, P. R., et al. (2011), The Community Climate System Model version 4,J. Clim., 24, 4973–4991, doi:10.1175/2011JCLI4083.1.

Hegerl, G. C., T. Crowley, M. Allen, W. T. Hyde, H. Pollack, J. Smerdonand E. Zorita (2007), Detection of human influence on a new 1500 yrclimate reconstruction, J. Climate, 20, 650–666.

Jansen, E., et al. (2007), Palaeoclimate, inClimate Change 2007: The PhysicalScience Basis—Contribution of Working Group I to the Fourth AssessmentReport of the Intergovernmental Panel on Climate Change, edited byS. Solomon et al., pp. 463–497, Cambridge Univ. Press, Cambridge, U.K.

Jones, P. D., M. New, D. E. Parker, S. Martin, and I. G. Rigor (1999),Surface air temperature and its change over the past 150 years, Rev.Geophys., 37, 173–199.

Kaufman, D. S., et al. (2009), Recent warming reverses long-term Arcticcooling, Science, 325(5945), 1236–1239.

Kinnard, C., C. M. Zdanowicz, D. A. Fisher, E. Isaksson, A. de Vernal, andL. G. Thompson (2011), Reconstructed changes in Arctic sea ice over thepast 1,450 years, Nature, 479(7374), 509–512.

Landrum, L., B. L. Otto-Bliesner, E. R. Wahl, A. Conley, P. J. Lawrence,N. Rosenbloom, andH. Teng (2013), Last millennium climate and its variabil-ity in CCSM4, J. Clim., 26, 1085–1111, doi:10.1175/JCLI-D-11-00326.1.

Langway, C. C., Jr., K. Osada, H. B. Clausen, C. U. Hammer, and H. Shoji(1988), An inter-hemispheric volcanic time-marker in ice cores fromGreenland and Antarctica, Ann. Glaciol., 10, 102–108.

Lavigne, F., et al. (2013), Source of the great A.D. 1257 mystery eruptionunveiled, Samalas volcano, Rinjani Volcanic Complex, Indonesia, Proc.Natl. Acad. Sci. U. S. A., doi:10.1073/pnas.1307520110, in press.

Mann, M. E., Z. Zhihua, M. K. Hughes, R. S. Bradley, S. K. Miller,S. Rutherford, and F. Ni (2008), Proxy-based reconstructions ofhemispheric and global surface temperature variations over the pasttwo millennia, Proc. Natl. Acad. Sci. U.S.A., 105(36), 13,252–13,257,doi:10.1073/pnas.0805721105.

Mann, M. E., Z. Zhang, S. Rutherford, R. S. Bradley, M. K. Hughes,D. Shindell, C. Ammann, G. Faluvegi, and F. Ni (2009), Global signaturesand dynamical origins of the Little Ice Age and Medieval ClimateAnomaly, Science, 326(5957), 1256–1260.

Mann, M. E., J. D. Fuentes, and S. Rutherford (2012), Underestimation ofvolcanic cooling in tree-ring-based reconstructions of hemispheric temper-atures, Nat. Geosci., 5, 202–205, doi:10.1038/NGEO1394.

Marsland, S. J., H. Haak, J. H. Jungclaus, M. Latif, and F. Roeske (2003),The Max-Planck-Institute global ocean/sea ice model with orthogonalcurvilinear coordinates, Ocean Modell., 5, 91–127.

Marti, O., et al. (2010), Key features of the IPSL ocean atmosphere modeland its sensitivity to atmospheric resolution, Clim. Dyn., 34, 1–26.

Meehl, G. A., et al. (2012), Climate system response to external forcings andclimate change projections in CCSM4, J. Climate, 25, 3661–3683,doi:10.1175/JCLI-D-11-00240.1.

Mignot, J., M. Khodri, C. Frankignoul, and J. Servonnat (2011), Volcanicimpact on the Atlantic Ocean over the last millennium, Clim. Past, 7,1439–1455, doi:10.5194/cp-7-1439-2011.


12,369

Miller, G. H., et al. (2010), Temperature and precipitation historyof the Arctic, Quat. Sci. Rev., 29(15–16), 1679–1715, doi:10.1016/j.quascirev.2010.03.001.

Miller, G. H., et al. (2012), Abrupt onset of the Little Ice Age triggered byvolcanism and sustained by sea-ice/ocean feedbacks, Geophys. Res.Lett., 39, L02708, doi:10.1029/2011GL050168.

Otterå, O. H., M. Bentsen, H. Drange, and L. Suo (2010), External forcing as ametronome for Atlantic multidecadal variability, Nat. Geosci., 3, 688–694.

Phipps, S. J. (2010), The CSIRO Mk3L climate system model v1.2,Technical Report No. 4, Antarctic Climate & Ecosystems CRC, Hobart,Tasmania, Australia, 122 pp., ISBN 978-1-921197-04-8.

Raddatz, T. J., C. H. Reick, W. Knorr, J. Kattge, E. Roeckner, R. Schnur,K. G. Schnitzler, P. Wetzel, and J. Jungclaus (2007), Will the tropicalland biosphere dominate the climate-carbon cycle feedback during thetwenty-first century?, Clim. Dyn., 29, 565–574, doi:10.1007/s00382-007-0247-8.

Robock, A. (2000), Volcanic eruptions and climate,Rev. Geophys., 38, 191–219.Schleussner, C. F., and G. Feulner (2013), A volcanically triggered regimeshift in the subpolar North Atlantic Ocean as a possible origin of theLittle Ice Age, Clim. Past, 9, 1321–1330, doi:10.5194/cp-9-1321-2013.

Schmidt, G. A., et al. (2006), Present day atmospheric simulations using GISSModel E: Comparison to in-situ, satellite and reanalysis data, J. Clim., 19,153–192.

Schmidt, G. A., et al. (2011), Climate forcing reconstructions for use inPMIP simulations of the last millennium (v1.0), Geosci. Model Dev., 4,33–45, doi:10.5194/gmd-4-33-2011.

Schneider, D. P., C. M. Ammann, B. L. Otto-Bliesner, and D. S. Kaufman(2009), Climate response to large, high-latitude and low-latitude volcaniceruptions in the Community Climate System Model, J. Geophys. Res.,114, D15101, doi:10.1029/2008JD011222.

Shindell, D. T., G. A. Schmidt, R. L. Miller, andM. E. Mann (2003), Volcanicand solar forcing of climate change during the preindustrial era, J. Clim., 16,4094–4107.

Steinhilber, F., J. Beer, and C. Frohlich (2009), Total solar irradiance during theHolocene, Geophys. Res. Lett., 36, L19704, doi:10.1029/2009GL040142.

Timmreck, C., S. J. Lorenz, T. J. Crowley, S. Kinne, T. J. Raddatz,M.A. Thomas,and J. H. Jungclaus (2009), Limited temperature response to the very largeAD 1258 volcanic eruption, Geophys. Res. Lett., 36, L21708, doi:10.1029/2009GL040083.

Timmreck, C., H.-F. Graf, S. J. Lorenz, U. Niemeier, D. Zanchettin, D. Matei,J. H. Jungclaus, and T. J. Crowley (2010), Aerosol size confines climateresponse to volcanic super-eruptions, Geophys. Res. Lett., 37, L24705,doi:10.1029/2010GL045464.

Toohey, M., K. Kruger, U. Niemeier, and C. Timmreck (2011), The influenceof eruption season on the global aerosol evolution and radiative impactof tropical volcanic eruptions, Atmos. Chem. Phys., 11, 12,351–12,367,doi:10.5194/acp-11-12351-2011.

Vieira, L. E. A., and S. Solanki (2010), Evolution of the solar magnetic fluxon time scales of years to millenia, Astron. Astrophys., 509, doi:10.1051/0004-6361/200913276.

Wang, Y. M., J. L. Lean, and N. R. Sheeley Jr. (2005), Modeling the Sun’smagnetic field and irradiance since 1713, Astrophys. J., 625, 522–538.

Watanabe, S., et al. (2010),MIROC-ESM2010:Model description and basicresults of CMIP5-20c3m experiments, Geosci. Model Devel., 4, 845–872,doi:10.5194/gmd-4-845-2011.

Wu, T., et al. (2013), Global carbon budgets simulated by the BeijingClimate Center Climate System Model for the last century, J. Geophys.Res. Atmos., 118, 4326–4347, doi:10.1002/jgrd.50320.

Zanchettin, D., O. Bothe, H. F. Graf, S. J. Lorenz, J. Luterbacher, C. Timmreck,and J. H. Jungclaus (2013), Background conditions influence the decadalclimate response to strong volcanic eruptions, J. Geophys. Atmos., 118,4090–4106, doi:10.1002/jgrd.50299.

Zhong, Y., G. H.Miller, B. Otto-Bliesner,M. Holland, D. Bailey, D. Schneider,and A. Geirsdottir (2010), Centennial-scale climate change from decadally-paced explosive volcanism: A coupled sea ice-ocean mechanism, Clim.Dyn., 37, 1–15, doi:10.1007/s00382-010-0967-z.


12,370

northern hemispheric cryosphere response to volcanic...

Documents